Adversarial Testing, was Thou shalt have no 1625

Douglas A. Gwyn

I disagree. The value of the model is precisely the fact that a single instance of the game does not have a dominant strategy. As you mentioned variations in the rules make a big difference.

Analyzing only the personal payoffs produced a clear preference for defection. Analyzing the full payoff matrix and thus recognizing that it is symmetric suggests that other strategies may be useful. When you add the requirement that all participants must buttume that their adversaries are rational, which requirement is not part of the clbuttic PD problem, but was part of the article you mentioned, then cooperation is the dominant strategy.

For other readers of this thread the logic is as follows:

1. Both participants will make their choice rationally.

2. The payoff matrix is symmetrical with respect to the participants so the conditions to be analyzed by the two players are identical.

3. Combining (1) and (2) we derive the expectation the all participants will make the same choice. I.e., that the result will come from the diagonal of the payoff matrix.

4. The best diagonal result is when both players cooperate, so that is the rational choice.

There a regress problem with this as to whether it is rational to buttume that the other player is rational, but I've already gone down the decidability rat hole once this month.

tj3